补偿模糊神经网络在磨矿控制中的应用研究

Application of Compensation Fuzzy Neural Networks in Grinding Control

针对典型两段式磨矿控制优化中系统机理复杂、影响因素多、难以建立精确的数学模型等诸多难点问题,采用补偿模糊神经网络对磨矿控制器进行设计,通过利用模糊控制对模糊信息的处理能力和神经网络强大的学习能力很好地解决了磨矿控制中非线性、难建模等问题。补偿模糊神经元的引入,能使网络从初始正确定义的模糊规则或者初始错误定义的模糊规则进行训练,使系统具有更高的容错性,系统更稳定。仿真结果表明,矿石粒度被很好地控制在了一个较理想的范围,证明了补偿模糊神经网络对磨矿控制的有效性和实用性。

Due to the complicated mechanism and many influence factors, an accurate mathematical model is dif- ficult to be established in typical two-stage grinding control. In this paper, Compensation Fuzzy Neural Networks （CFNN） is used to design the grinding controller. By using the processing ability of fuzzy control for fuzzy information and the strong learning ability of neural network, nonlinear and modeling difficulty in grinding control are well solved. By introducing compensatory fuzzy neural cells, the network with correct or erroneous fuzzy rules defined initially can be trained to possess higher fault tolerance and stability. The result of simulation shows that the ore granularity can be well controlled in an ideal scope. It proves the effectiveness and practicality of CFNN to the grinding control.